genetic estimates of population structure and gene flow limitations lessons and new directions

PERSPECTIVES

Genetic estimates of population structure
and gene flow: limitations, lessons and
new directions
J.L. Bossart
D. Pashley Prowell
indirect methods using genetic markers are the primary measure of gene flow levels
among interbreeding populations. Results from studies employing these methods are
often ambiguous and open to multiple interpretation. This is primarily due to low
resolution of molecular markers and low precision of model-based estimates. Studies
of paternity, kinship and phylogeography generate the most reliable results. Future
studies should employ more powerful analytical methods, analyse loci independently,
and attempt to distinguish confounding contributions of vicariance to isolation-bydistance studies. Moreover, direct assessment of movement remains the most valid
approach to many key issues in population biology.
J.L. Bossart and D.P. Prowell are at 402 Life Sciences Building,
Dept of Entomology, Louisiana State University,
Baton Rouge, LA 70803,USA.

tudies of gene flow are integral to interS
pretation of microevolutionary patterns

and geographic structure. Through such
studies, we strive to gain insights into eve
lutionary independence and potential for
population diversification, differentiation,
and ultimately speciation. Indirect methods, which infer gene flow from genetic
data on measurements of population structure, are relied upon almost exclusively to
calculate this important parameter (see
Box 1 for definitions of relevant terms).
This widespread popularity reflects the
ease with which genetic data can generally be collected and analysedl, and the
belief that such estimates may be more
valid than those obtained through direct
studies of movement* because they give a
temporal perspective. However, many indirect estimates of gene flow may be of
limited interpretive value and reveal little
about the degree of genetic cohesion and
evolutionary potential of contemporary
populations, for two reasons: (1) molecular
markers appear to lack the resolving power
required to distinguish contemporary patterns of gene flow with current analytical

methods; (2) limitations imposed by conventional population genetics models lead
to low precision of gene flow estimates
and a general lack of biological realism.
Others have emphasized the difficulties associated with interpreting patterns
of molecular variation and model-based
estimates (for example, Ref. 3). Yet these
cautions have not been widely embraced
by the scientific community. Conclusions
often are drawn about genetic structure
and levels of gene flow even though there
are multiple, equally viable interpretations

202

Copyright

0 1998, Elsevier

Science

of results. Moreover, these methods are
advocated for use in ecological studies4,
such as population demography, metapopulation processes, and interspecific
interactions, but there is much uncertainty
about the success of molecular methods
in ecological time scales. Indeed, presentday patterns of genetic structure may often reflect effects of Pleistocene glaciations
and post-glacial range expansion&s.

Problems with interpretation of
indirect estimates
Most population genetic studies using
indirect methods seek to establish levels
of gene flow. Statistics that estimate this
parameter generally include variance
among populations in allele frequencies
(Fsr, Wright’; GsT, Neia; and 0, Weir and
Cockerhams; and their analogues, e.g. R,,
Slatkinl”) and Nm,the effective number of
migrants per generation’. Values are interpreted as indications of relative levels of
gene flow among populations. For instance,

high Nm and low FsT indicate high migration. Additional analyses are often conducted based on spatial relationships predicted by isolation-by-distance models.
Correlations between geographic distance
and genetic distance are calculated, and
geographic patterns visualized by clustering algorithms. Three possible results
obtained from these studies are: (1) no
genetic structure or differentiation among
populations; (2) significant genetic structure but no geographic pattern to the
structure; or (3) significant genetic differentiation and geographically structured
populations. Difficulties arise in the interpretation of all of these results.

Ltd. All rights

reserved.

016!k5347/98/$19.00

The first case can produce ambiguous results because there are multiple
explanations for homogeneity that cannot
be readily distinguished. These include
current gene flow among populations,

balancing selection on markers, and lack
of resolution of markers (i.e. retention
of shared ancestral polymorphisms). Scientists generally favor one of these explanations over the others. Occasionally,
a follow-up study with a second marker
detects significant heterogeneity among
populations and implicates lack of resolution for the first marker (for example, Zinku
detected no pattern in fox sparrows with
allozymes but later detected a pattern
with mtDNAl2). Unless lack of resolution is
documented, distinguishing among explanations is conjectural.
Results are also ambiguous in the second case where population structure is
detected by a significant FsT,yet no correlation with geography or geographic structuring is observed. Despite this result, estimates of Nmoften are taken at face value as
the approximate number of migrants moving among populations. However, these
estimates are not necessarily indicative of
gene flow for at least three reasons. First,
if these measures provide meaningful estimates of gene flow, there should be hierarchical structure that is associated with
geography. Second, significant discrepancies among the various statistics used to
partition genetic variation suggest problems in general with the methods and
their underlying assumptions. For example, Fsr, Gs,, and private allele analyses
do not always generate comparable Nm

valueslsJ4. Third, demographic instability
or heterogeneity among populations can
cause loose correlations between I?& and
Nm (Ref. 15). FsT may fluctuate over time
and space, resulting in a poor estimation
of genetic structure and gene flow.
The third outcome, significant genetic
and geographic structure, can be ambiguous because patterns can result from two
distinct processes. Structure may result
when vicariant events restrict gene flow
among a subset of populations (i.e. geographic isolation). Sharp discontinuities
in gene frequencies are usually evident on
either side of the vicariant event with genetic homogeneity among populations on
the same side. Alternatively, gene flow
levels among populations may vary and
follow an isolation-by-distance model with
neighboring populations exchanging more
migrants than distant ones, lsolation-bydistance theoretically results in a direct
relationship between genetic distance
and geographic distance. In reality, these

two alternatives cannot be easily distinguished because the scatter of points in
the distance regression is usually large.
Furthermore, it is likely that many studies

PII: SO169-5347(97)012&I-6

TREE

vol.

1.3, no.

5 May

I998

PERSPECTIVES
of isolation-by-distance are, in fact, confounded by vicariance because vicariance
is more likely to be detected with increasing geographic distance. Irrespective of
the term used to define such isolated taxa,

one is not studying levels of gene flow in
contemporary populations (i.e. Nm has
been zero over some period of evolutionary time), but instead, geographic variation among allopatric taxa (i.e. phylogeography). There are a few studies that
appear to support isolation-by-distance,
usually in very ‘sedentary’ organisms (e.g.
Refs 14,16,17). However, in these studies, it is not possible to discount the fact
that samples contributing to a significant
genetic-geographic correlation may be
completely isolated.
Many of these interpretational problems have been recognized previouslysJ*,
and have fostered an expanding interest
in the development of high resolution molecular markers. Population genetic studies have moved beyond allozymes and now
routinely include data from DNA regions.
In general, allozymes evolve at a slower
rate than mtDNA and nuclear DNA such
as microsatellitesisJ0 and rarely provide
enough resolution to track contemporary
processes and assess gene flow levels. Allo
zymes are good at detecting vicariancel.
Although DNA markers evolve at faster

rates and thus potentially reveal more
recent evolutionary events, their ability to
resolve levels of gene flow remains uncertain. In a cursory review of the recent literature, we found little evidence that DNA
markers give better estimates of gene flow
(Table 1). In summary, allozymes and DNA
markers produced similar results except
in cases where a vicariant event was suggested by the DNA but not the allozymes
(for example, Ref. 22). The challenge thus
remains to separate genetic history from
present-day patterns. Moreover, although
indirect methods clearly reveal the history of gene frequency distributions up to
some ‘recent’ point, the precise location
of that point is unknown, The presumed
advantage of molecular approaches is their
temporal perspective on genetic structure, but this is only an advantage to the
extent the evolutionary time scale of this
temporal perspective can be determined.

Problems with population genetic
models

A second issue regarding indirect methods concerns limitations imposed by population genetics models (for example, Ref.
29, p. 863). These limitations are sufficiently
stringent that it is difficult to have confidence in estimates obtained. As Lewontins
cautioned, model-based estimates of evolutionary parameters, such as gene flow,
will almost never be valid because parameters that define these models are spatially
TREE

uol.

13, no.

5May

1998

Box 1. Definitions

of terms

Contemporary

gene flow: present-day pattern of movement of genes (via individuals) among interbreeding
populations of a species.
Direct measures of movement:
use of genetic (constructed)
or physical markers that allow tracking of
dispersal of individuals among populations.
Indirect estimates of gene flow: movement of genes (via individuals) inferred from genetic data on
measurement
of population structure.
isolation-by-distance:
decreasing gene flow and thus genetic similarity between populations with increasing
geographrc distance.
Marker resolution: relative power to reveal meaningful patterns of genetic structure and track contemporary
gene flow; largely a function of mutation rate.
Vicariance: historical process of range fragmentation
that leads to geographic isolation of populations
within an ancestral species.

Table 1. Comparison of congruence and resolution of allozyme and DNA
markers used to assess gene flow levels in some recent studies
Geneticgeographic
correlation

Geographic
pattern

Taxon

Marker

Significant
heterogeneity

Anopheline mosquito21
(Anopheles gambiae)

Allozymes
Microsatellites

N?
N?

NA
NA

NA
NA

American oysterz2
(Crassoshea virginica)

Allozymes
mtDNA
Nuclear RFLPs

N
Y
Y

NA
NA
NA

N
Y*
Y*

Sea beetz3
(Beta vu/gafis maritima)

Allozymes
Nuclear RFLPs

Y
Y

N
Y

NA
NA

Atlantic salmo@
(Salmo salar )

Allozymes
Microsatellites

Y
Y

Y
Y

Y*
Y*

Atlantrc cod25
(Gadus morhua)

Allozymes
Nuclear RFLPs

Y
Y

Y
Y

Y*
Y*

Cyprinid minnow76
(Tiaroga cobitis)

Allozymes
mtDNA

Y
Y

NA
NA

Y*/N
Ye/N

Cyprinid minnows*’
(Meda fulgida)

Allozymes
mtDNA

Y
Y

NA
NA

Y*
Y’

Fox sparrow
(Passerella

Allozymeslf
mtDNAl2

N
Y

NA
NA

N
Y*

Allozymes*’
mtDNA2s

Y
Y

NA
NA

Y*
Y*

iliaca)

Threespine stickleback
(Gasterosteus
aculeatus)
N=No, Y=Yes; NA=value
Y/N=some
Inconsistencies
heterogeneity not tested.

not calculated;
in geographic

* =vicariance
probably responsible
pattern; N?=some
IOCI significant,

for observed pattern;
significance of overall

I

1

Box 2. Abiotic and biotic factors that promote genetic and demographic
instability and non-uniformity across a species’ range
l

l
l
l

Weather-related
effects, both catastrophic
deterministic
Habitat patch location. e.g. edge or central
Variable patch size
Variable distance between habitat patches

and

and temporally variable and too sensitive
to forces of evolution. More generally,
population genetics models ignore the
complexity that defines biological systernGO. Conventional models assume that
population structure and demographic parameters such as population size and dispersal rates are uniform and constant over
space and time. Assumptions of demographic and genetic equilibrium and uniformity are unrealistic and violated as a

l
l

l
l

Variable resource availability and quality
Rates of gene flow that evolve in response
habitat patch attributes
Variable brood numbers and size
Interactions with other organisms

to

matter of course. Demographic parameters
are a function of abiotic and biotic factors
that regularly change (Box 2) often in unpredictable ways (e.g. Ref. 31). These de
terministic and stochastic elements and
their interactions routinely perturb and
introducevariability into natural systems,
as commonly evidenced in the literature.
Population genetic studies, however, only
rarely address basic model assumptions of
population structure and equilibrium with

203

PERSPECTIVES

Box 3. Potential sources of error and ambiguity surrounding
genetic structure and gene flow
l
l
l
l
l
l
l
l
l
l

Choice of analytical method used to partition genetic variation
Choice of model to estimate gene flow
Choice of marker
Inherent error associated with second order f-statistics
Violations of assumptions of analytical methods, genetic models, or both
Variation among locus specific estimates and averaging over loci
Variation among allele specific estimates and averaging over alleles
Inadequate or improper sampling of species’ range
Confounding of contemporary patterns with historical associations
Cumulative averages that confound gene flow and other evolutionary determinants
among populations (e.g. demographic fluctuations and genetic bottlenecks)

respect to populations of interest. Moreover, discussions rarely include explanations of how violations of these assump
tions affect results.
From a purely statistical perspective,
effects of genetic and demographic nonuniformity and stochasticity on estimates
of genetic structure and model-based parameters include scale dependence of FsT
(e.g. Refs 14,32), fluctuation of FST over
space and time (e.g. Ref. 15), wide variance of FSTvalues among loci (e.g. Ref. 33),
and a loose correlation between FST and
Nm (Ref. 15). Thus, estimates of Nm will
have large standard errors. ‘Snapshot’ estimates assessed at one (or a few) points in
space and time will be unlikely to reveal
actual levels of interpopulation genetic exchange. Note that population genetic estimates based on F-statistics inherently
have large varianceszY,irrespective of variance attributable to genetic and demographic factors.
demographic
More
importantly,
instability determines, in part, the evolution of genetic variation among populations3°s34,35.36.
Fluctuating or novel conditions of genetic variance that occur as a
consequence may be integral to population
diversification and speciation, and temporarily provide conditions conducive to
kin selection or shifting balance evolution15.“0X33835.
Conventional indirect approaches implicitly assume that gene frequency distributions among populations
are caused by gene flow (e.g. Ref. 29).
Contributions of other evolutionary forces
to genetic structure are not distinguished
and estimates obtained are interpreted as
weighted averages of effective ‘gene flow’
over time. Such cumulative weighted averages however may provide few insights
into the issues we would most like to address as evolutionary biologists, that is,
the microevolutionary processes that promote or hinder genetic differentiation and
diversification of populations.
As Strong37 and others have previously
emphasized ‘ . . .mathematical theory of
equilibrium and stability has proven disappointingly sterile for ecology’. Dispersal
and rates of gene flow will likely be unique
204

estimates of

of gene frequencies

characteristics of different populations
and most appropriately viewed as a dynamic, interconnected network of differential values. Moreover, to the extent that
individuals differ genetically in their pre
pensity to disperse and wanders*, attributes of different habitat patches will selectively act on this underlying variation
such that rates of gene flow will ‘evolve’
like any other quantitative trait. Just as
the evolution and spread of adaptive traits
can generally not be assessed from studies of population means, but rather require
detailed analysis of (co)variance structure,
the importance of gene flow as a determinant of genetic pattern across environmental landscapes will be difficult to evaluate from cumulative weighted averages.
Future studies
After more than 25 years of using molecular methods to estimate genetic structure and gene flow, enough data have accumulated to assesswhere the markers have
succeeded and where they have not. Their
most successful use is in phylogeography
and species level analyses1t3g.Allozymes are
generally sufficient for detecting isolation
caused by vicariant events, but in some
groups higher resolution DNA markers are
required. Some of the more variable markers such as microsatellites may be excellent
tools for establishing kinship relationships
among related individuals (e.g. Ref. 16) and
even allozymes have proven useful in this
respect (e.g. Ref. 40). There appear to be
many fewer successes in areas between
these two extremes, that is, evaluating levels of gene flow among interbreeding populations and in population demographic
studies. Here, efforts are undermined by
the many sources of error and ambiguity associated with indirect approaches (Box 3).
Despite shortcomings summarized
here, genetic approaches are a useful starting point for evolutionary studies when
results are properly analysed statistically
and interpreted critically, and when the
question of interest is an appropriate ap
plication of their use. In this regard, we
offer four specific recommendations for
consideration.

Analyses of population structure rely
almost exclusively on antiquated methodology. From the beginning of 1996 to mid
1997, 21 articles published in the journal
Evolution used indirect methods to measure population structure. Of these, almost
all based interpretation on F-statistics
analysis (18 of 21 studies) and allozyme
markers (15 of 21 relied solely on allozymes). Although most population geneticists acknowledge that Nm imparts minimal information, this value is still widely
reported (14 of 21 studies) and often literally interpreted. Theoretical population
genetics methodology has sufficiently
progressed that we should be hesitant to
base measures of genetic structure on Fstatistics analysis, especially of allozyme
variation, and simple genetic models. More
realistic models and analytical methods
are available and should be evaluated
for their precision and statistical power
(e.g. Refs 41,42). Many of these new methods rely upon DNA data and genealogical
analysis of haplotypes or alleles and have
increased capacity for distinguishing
genetic/spatial associations including vicariance, isolation-by-distance, and range
expansion.
Second, variance among single locus
estimates of genetic structure is common.
This suggests that either uninformative
loci have been included in the analysis (i.e.
those exhibiting low levels of polymorphism and thus unlikely to reveal genetic
structure) or that basic assumptions of
genetic equilibrium or marker neutrality
or both have been violated7*3*. Our current
approach of combining data across discrepant loci introduces error in the estimation procedure. More importantly, when
locus-to-locus variance cannot be explained by disparate levels of allelic diversity, estimates obtained will be biologically
misleading. Such estimates confound gene
flow with other evolutionary determinants
of genetic variance among populations and
thus do not take full advantage of the information we have available. If our goal is
to distinguish relative effects of different
evolutionary forces in population diversification, discrepant loci should be emphasized for their power to reveal underlying
mechanisms. Compatibility assessments
across loci would be more informative42
and locus-specific estimates should routinely be reported and interpreted in the
context of allelic diversity, model assumptions, locus-specific evolutionary forces,
and observed patterns of genetic structure.
Our third suggestion concerns analysis of isolation-by-distance. Such studies
generally employ regression analysis of
data derived from estimates of genetic distance (or gene flow) among pairwise population contrasts. Effects of single populations or clusters of populations are rarely
TREE

uol.

1.7, no.

5 May

1998

PERSPECTIVES
distinguished. Statistical significance tests
of each pairwise value should routinely be
conducted to assesswhether certain populations are largely responsible for isolationby-distance correlations. If statistically
significant values are associated with a single population or cluster of populations,
vicariance should be discussed as a viable
explanation. If significance is scattered
throughout population pairs, a stronger
case can be made for isolation-by-distance.
Our final recommendation concerns
direct studies of movement. Even our most
enthusiastic supporters of genetic approaches have long advocated the key
role such studies playl8. Yet data from direct observational studies have accumulated at a much slower rate than data from
indirect studies. Admittedly, direct studies
of movement are logistically difficult and
suffer from their own set of limitationsl8.
The most severe of which is that such
studies may tell us little about the importance of gene flow over evolutionary time.
However, direct observational studies of
movement can provide data on the degree
of spatial and temporal heterogeneity of
movement patterns across a species’
range, the impact of environmental attributes such as habitat patchiness and resource quality on movement patterns, and
the extent of quantitative genetic variation
for propensity to stay or wander. Moreover, direct assessment of movement
among contemporary populations remains
the only valid approach to the study and
interpretation of gene flow in an ecological context. Technological advances will
undoubtedly facilitate such studies in the
future. The use of uniquely marked or
tagged individuals tracked by satellite or
radar will allow unbiased assessments of
direct movement over short and long distances for organisms amenable to such
tags (e.g. Ref. 2). Indeed, the use of genetic
tags to track the movement of marine organisms has proven to be a successful
arena for applying genetic approaches to
the study of movement among presentday populations43,M.
In summary, our understanding of
gene flow will progress more significantly
through candid discussion of limitations,
more refined statistical analysis, and a better balance between indirect and direct approaches. Our reliance on easy to apply,
conventional indirect methods to study
population structure and gene flow is no
longer advancing the field. Technological
advances in DNA analysis over the past
decade have brought about a new era of
data gathering that has potential for enhanced resolution. The challenge now is
to develop and implement analytical methods that keep pace with technology and
allow more precise estimates of gene flow
in contemporary time.
TREE

vol.

13, no.

5 May

1998

Acknowledgements

We thank Chris Carlton, David Foltz,
Joseph Neigel, Robert Zink and three
anonymqus reviewers for their
comments on early drafts of the
manuscript. Special thanks to
John Thompson for an introduction
at a meeting that led to formalization
of our ideas. Our research is supported
by an NSF/Louisiana Education Quality
Support grant #[ 199%1997)~ADP-02 and
USDA/CRGO 93-37302-9128.
References
1 Avise, J.C. (1994) Molecular Markers, Natural
History and Evolution, Chapman &Hall
2 Koenig, W.D., van Vuren, D. and Hooge. P.N.
(1997) Detectability,
philopatty
and the
distribution
of dispersal distances in
vertebrates,
Trends Ecol. Evol. 11,
514-517
3 Lewontin, R.C. (1985) Population
genetics, in
Evolution: Essays in Honour of John Maynard
Smith (Greenwood,
P.J., Harvey, P.H. and
Slatkin. M., eds), pp. 3-18, Cambridge
University
Press
4 Roderick. G.K. (1996) Geographic
structure
of insect populations:
gene flow,
phylogeography
and their uses, Annu. Rev.
Entomol. 41, 325-352
5 Larson, A., Wake, D.B. and Yanev. K.P. (1984)
Measuring
gene flow among populations
having high levels of genetic fragmentation,
Genetics 106,293-308
6 Merila, J., Bjorklund,
M. and Baker, A.J. (1997)
Historical
demography
and present day
population
structure
of the greenfinch,
Curduelis
chloris
- an analysis of mtLINA
control-region
sequences, Evolution 51,
946-956
7 Wright, S. (1951) The genetical structure of
populations,
Ann. Eugen. 15,323-354
8 Nei, M. (1973) Analysis of gene diversity
in
subdivided
populations,
Proc. Natl. Acad. Sci.
U. S. A. 70,3321-3323
9 Weir, B.S. and Cockerham,
C.C. (1984)
Estimating Fstatistics
for the analysis of
population
structure, Evolution 38,
1358-1370
10 Slatkin, M. (1995) A measure of population
subdivision
based on microsatellite
allele
frequencies,
Genetics 139,457-462
11 Zink, R.M. (1986) Patterns and evolutionary
significance
of geographic
variation
in the
Schistacea group of the fox sparrow
(Passerella
ifiaca),
Ornithol. Monogr. 40
12 Zink. R.M. (1994) The geography
of
mitochondrial
DNA variation,
population
structure,
hybridization
and species limits in
the Fox Sparrow (Passerella
iliaca),
Evolution 48, 96-l 11
13 Slatkin. M. and Barton, N. (1989)
A comparison
of three methods for
estimating
average levels of gene flow,
Evolutron 43,1349-1368
14 Husband, B.C. and Barrett, S.C.H. (1995)
Estimates of gene flow in Eichhornio
paniculata
(Pontederiaceae):
effects of
range substructure,
Heredity 75549-560
15 Whitlock, M.C. (1992) Temporal fluctuations
in demographic
parameters
and the genetic
variance among populations,
Evolutron 46,
608-615

16 Peterson, M.A. (1996) Long-distance
gene
flow in the sedentary
butterfly,
Euphilotes
enoptes
(Lepidoptera:
Lycaenidae),
Evolution
50, 1990-1999
17 Hellberg, M.E. (1996) Dependence
of gene
flow on geographic
distance in two solitary
corals with different
larval dispersal
capabilities,
Evolution 50, 1167-l 175
18 Slatkin, M. (1985) Gene flow in natural
populations,
Annu. Rev. Ecol. Syst. 16,
393-430
19 Mitton, J. (1994) Molecular
approaches
to
population
biology, Annu. Rev. Ecol. Syst. 25,
45-69
20 Queller, DC, Strassman, J. and Hughes. CR.
(1993) Microsatellites
and kinship, Trends
Ecol. Evol. 8,285-288
21 Lehmann, T. et al. (1996) Genetic
differentiation
of Anopheles
gambiae
populations
from East and West Africa;
comparison
of microsatellite
and allozyme
loci, Heredity 77,192-208
22 Karl, S.A. and Avise, J.C. (1992) Balancing
selection at allozyme loci in oysters:
implication
from nuclear RFLP’s, Science 256,
100-102
23 Raybould, A.F.. Mogg, R.J. and Clarke, R.T.
(1996) The genetic structure
of Beta vulgaris
ssp. Armitima
(sea beet) populations:
RFLF%
and isozymes show different patterns of
gene flow, Heredily 77,245-250
24 Sgnchez, J.A. et al. (1996) Protein and
microsatellite
single locus variability
in
Sulmo
salar
L (Atlantic salmon), Heredity 77,
423-432
25 Pogson, G.H., Mesa, K.A. and Boutilier, R.G.
(1995) Genetic population
structure and gene
flow in the Atlantic Cod Gadus
morhus:
a comparison
of allozyme and nuclear RFLP
loci, Genetics 139, 375-387
26 Tibbets, C.A. and Dowling, T.E. (1996) Effects
of intrinsic and extrinsic factors on
population
fiagmetitation
in three species of
North American minnows fleleostei:
Cyprinidae),
Evolution 50,1280-1292
27 Haglund, T.R., Buth, D.G. and Lawson, R.
(1992) Allozyme
variation
and phylogenetic
relationships
of Asian, North American,
and
European populations
of the threespine
stickleback,
Gasterosteus
orufeatus,
Copeia
1992,432-443
28 Orti. G. et al. (1994) Global survey of
mitochondrial
DNA sequences in Threespine
Stickleback:
evidence for recent migrations,
Evolution 48, 608-622
29 Cockerham,
CC. and Weir, B.S. (1993)
Estimation of gene flow from F-statistics,
Evolution 47, 855-863
30 Slatkin, M. (1989) Population
structure
and
evolutionary
progress, Cenome 31,196-202
31 Andrewartha,
H.G. and Birch, L.C. (1954)
The Distribution and Abundance of Animals,
University
of Chicago Press
32 Johannsen. J.. Veith, M. and Seitz, A. (1996)
Structure
of the butterfly
Melitaea
didyma
(Nymphalidae)
along a northern
distribution
range border, Mol. Ecol. $259-267
33 Levin, D.A. (1988) Consequences
of stochastic
elements in plant migration,
Am. Nat. 132,
643-651
34 Wright. S. (1948) On the roles of directed
and random changes in gene frequency
in
the genetics of populations,
Evolution 2,
279-294

205

PERSPECTIVES
Movement
(Swingland, 1.R.and
Greenwood, P.J.,eds), pp. 102-115, Clarendon
Press
39 Berlocher, S.H.(1984) Insect molecular

35 Endler, J. (1977) Geographical
Variation,
Speciation and Cl&s, Princeton University
Press
36 Loveless, M.D. and Hamrick, J.L. (1984)
Ekological determinants
of genetic structure
in plant populations,
Annu. Rev. Ecol. Syst.

systematics,

15.

65-95
37 Strong, D.R. (1986) Density-vague
population
change, Trends Ecol. Evol. 1, 39-42
38 Swingland, I.R. (1983) Intraspecific
differences
in movement,
in The Ecology of

‘Lamarckian’
evolution

Annu.

Rev.

Entomol.

41

al. (1996) Molecular
evidence
for kin groups in the absence of large-scale
genetic differentiation
in a migratory
bird,

Ecol. (in press)
43 Bowen, B.W. (1995) Tracking

Evolution

44 Palsb011,P.J. el al. (1997) Genetic tagging of

mechanisms

of

in darwinian

amarckism’
‘darwinism’ are traL
ditionally seen as alternative theories
trying to account for evolutionary change

General selection theory makes no assumptions about the origin of heritable
variation. It maintains that evolution by
natural selection will occur in any system
(Box 1). The verdict of history is that
Lamarck got it wrong- evolutionary change with entities manifesting the properties
does not occur through the inheritance of of multiplication, heredity and heritable
acquired characters. Acquired characters variation affecting reproductive successl.
are the outcome of instructive processes, In the current version of biological darwinsuch as those seen in embryonic induc- ism, it is assumed that information is digition, transcriptional regulation, and learn- tal and encoded in DNA base sequences,
ing, all of which involve highly specific that multiplication of information occurs
and usually adaptive responses to factors through DNA replication, and that variexternal to the responding system. The ation, which is generated by mutation and
inheritance of the outcomes of instructive
recombination, is random with respect to
the selecting environment and the develprocesses is deemed to be impossible.
Adaptive evolutionary change is assumed opmental history of the organism and the
to be based on darwinian (or more accu- lineage. However, this version of evolution
rately neo-darwinian) evolution in which - ‘genie neodarwinism’ - is incomplete: it
guidance comes exclusively from selective gives natural selection an exclusive deterprocesses. The production and nature of ministic role in the evolution of all conceivheritable variation is assumed to be un- able complex adaptations, but until reinformed by the environment or by previ- cently it has had rather little to say about
ous history. The future is open-ended, de- the evolution of new systems for acquiring,
termined solely by the contingencies of storing and transferring information, and
life. It is neither foretold nor intimated,
even less about the evolutionary effects
and

b 1998, Elsevier

Science

Ltd. All rights

with genetic

50, 924-930

Neigel, J.E. (1997) A comparison

Eva Jablonka and Marion Lamb are at the Cohn Institute for the History and Philosophy of
Science and Ideas, Tel-Aviv University, Tel-Aviv 69978, Israel;
Eytan Avitai is at the Dept of Natural Sciences,David Yelin Teachers College,
POB 3578, Jerusalem, Israel;
Eva Jablonka is also at the Wissenschaftskolleg zu ‘Berlin,
Wallotstrasse 19, D-14193Berlin, Germany (jablonkatiiko-berlin.de).

Copyright

of phylogeographic
data: testing hypotheses
about gene flow and population
history, Mol.

29,

Since the Modern Synthesis, evolutionary biologists have assumedthat the genetic
system is the sole provider of heritable variation, and that the generation of heritable
variation is largely independent of environmental changes. However, adaptive
mutation, eplgenetlc inheritance, behavioural inheritance through social learning,
and language-basedinformation transmissionhave properties that allow the
inheritance of induced or learnt characters. The role of induced heritable variation in
evolution therefore needsto be reconsidered,and the evolution of the systems that
produce induced variation needsto be studied.

206

Syst. 28, 105-128
42 Templeton, A.R. (1997) Nested clade analyses

403-434
40 Friesen, V.L. et

Eva Jablonka
Marion 1. Lamb
Eytan Avital

6

alternative
strategies for estimating
gene
flow from genetic markers,
Annu. Rev. Ecol.

Animal

reserved.

0169~5347/98/$19.00

humpback

markers,
whales,

marine

Bioscience
Nature

turtles

45,528-534

388,767-769

of such systems once they are in place.
Natural selection leads not only to the evolution of eyes, wings, and sonars, but also
to the evolution of new evolutionary rules.
Many of these rules undermine the assumption that variation is random. Mechanisms allowing the inheritance of acquired
characters have evolved several times
during the history of life, and understanding their evolution is crucial to understanding the transitions to new levels of
individuality*-4.

Evolved ‘lamarckian’ heredity
systems

The heredity systems that we consider
here are all complex mechanisms for the
acquisition, storage and transfer of information. All evolved through natural selection, but they differ from each other in the
type of information they transmit, in their
evolutionary history, and in their evolutionary effects. They include adaptive mutational systems involving non-random
changes in DNA, cellular heredity systems
in which information is acquired and transmitted through intracellular structures and
biochemical mechanisms, the transfer of
patterns of behavlour through social learning coupled with certain types of social organization, and the transmission of information using symbolic languages. All of
these systems allow certain outcomes of
the interaction between the organism and
its environment to be incorporated into
and maintained within the informationcarrying system, and the information to
be transmitted to future generations. All
therefore allow the inheritance of acquired
or learnt characters.
Adaptive mutational systems:
the intelligentgenome

In the genetic inheritance system, information for making RNA and proteins is
stored in DNA base sequences: an elaborate enzyme system enables this information to be replicated and transmitted
to the next generation. Physico-chemical
damage to the DNA and errors occurring
during its replication can be removed by a
battery of repair processes. Errors that
remain, and sequence changes that are

PII: SOlW5347(98)01344-S

TREE

uol.

1.7, no.

5 May

19.98